Medical Analytics Based on Artificial Neural Networks Using Cognitive Internet of Things

Abstract

This research proposes machine learning algorithms in conjunction with cognitive-based networking as a remote patient monitoring framework for accurately predicting disease state and disease parameters from remotely monitored and measured patient biometric and biomedical signals. This system would facilitate doctors and clinicians by providing hospitals machine learning-based predictive clinical decision support systems to remotely monitor patients and their diseases. In this proposed work, a cognitive radio (CR) network is simulated for optimization of spectrum sensing and energy detection. Further, two effective classification methods are evaluated on remotely measured physiological parameters, such as blood pressure and heart rate, of patients with two types of diseases—chronic kidney disease and heart disease. First, a support vector machine (SVM) model was trained on a heart disease dataset with inputs and binary targets. The disease parameter correlations between blood pressure and age, heart rate, and blood glucose level results were plotted and their relationships were modeled using SVM. Second, the artificial neural network (ANN) algorithm was employed for the detection of disease state with the two types of disease datasets—heart disease and chronic kidney diagnosis. With SVM, the accuracy was around 60% for heart disease and 84% for chronic kidney disease patients. The percentage of accurately categorized patients with ANN was observed to be 95% overall in estimate for heart disease and 93% overall in estimate for chronic kidney disease. ANN is more accurate and recommended for predictive modeling of patient data in the proposed cognitive IoT remote patient monitoring system.